ADVANCED MANUFACTURING SYSTEMS

ASSIGNMENT NO 1

Submitted to:

Dr. Amar-Ul-Hassan

Submitted by:

Faizan Ahmad

20-ME-67

DEPARTMENT OF MECHANICAL ENGINEERING UNIVERSITY

Generate a 10×8 (10 machines and 8 parts) machine-part incidence matrix having at least 50

operations and then by applying any approach i.e. CIA or ROC block-diagonalize this matrix and

calculate its Grouping Efficacy (GE). Analyze the outcome and develop an argument that whether

this solution can be further improved?

Solution

Rank Order Clustering Algorithm

10 machines and 8 parts machine-part incidence matrix having 55 operations and applying ROC
block-diagonalize this matrix and calculate its Grouping Efficiency (GE) at the end.

P1 P2 P3 P4 P5 P6 P7 P8

1 1 1 1 1 1

1 1 1 1
1 1 1 1 1 1 1

1 1 1 1 1 1

1 1 1 1 1

1 1 1 1 1

1 1 1 1 1

1 1 1 1 1 1

1 1 1 1 1

1 1 1 1 1 1

Assign binary weight and calculate a decimal weight for each row and column and rank the rows

in order of decreasing decimal weight values. Repeat steps 1 and 2 for each column and continue

preceding steps until there is no change in the position of each element in the row and the column.
 128 64 32 16 8 4 2 1 Binary Weight

Manufactures/Parts P1 P2 P3 P4 P5 P6 P7 P8

M1 1 1 1 1 1 1 237
M2 1 1 1 1 212
M3 1 1 1 1 1 1 1 191
M4 1 1 1 1 1 1 246
M5 1 1 1 1 1 157
M6 1 1 1 1 1 94
M7 1 1 1 1 1 167
M8 1 1 1 1 1 1 95
M9 1 1 1 1 1 143
M10 1 1 1 1 1 1 215

Binary 128 64 32 16 8 4 2 1
Weight
Manufacture P1 P2 P3 P4 P5 P6 P7 P8
s/Parts
512 M4 1 1 1 1 1 1 246
256 M1 1 1 1 1 1 1 237
128 M10 1 1 1 1 1 1 215
64 M2 1 1 1 1 212
32 M3 1 1 1 1 1 1 1 191
16 M7 1 1 1 1 1 167
8 M5 1 1 1 1 1 157
4 M9 1 1 1 1 1 143
2 M8 1 1 1 1 1 1 95
1 M6 1 1 1 1 1 94
1020 963 816 747 303 102 695 446 Decim
3 al
Binary
128 64 32 16 8 4 2 1
Weight
Manufacture P6 P1 P2 P3 P4 P7 P8 P5
s/Parts
512 M4 1 1 1 1 1 1 246
256 M1 1 1 1 1 1 1 237
128 M10 1 1 1 1 1 1 215
64 M2 1 1 1 1 212
32 M3 1 1 1 1 1 1 1 191
16 M7 1 1 1 1 1 167
8 M5 1 1 1 1 1 157
4 M9 1 1 1 1 1 143
2 M8 1 1 1 1 1 1 95
1 M6 1 1 1 1 1 94
1023 1020 963 816 747 695 446 303 Decim
al

Binary
128 64 32 16 8 4 2 1
Weight
Manufacture P6 P1 P2 P3 P4 P7 P8 P5
s/Parts
512 M4 1 1 1 1 1 1 246
256 M1 1 1 1 1 1 1 237
128 M10 1 1 1 1 1 1 215
64 M2 1 1 1 1 212
32 M3 1 1 1 1 1 1 1 191
16 M7 1 1 1 1 1 167
8 M5 1 1 1 1 1 157
4 M9 1 1 1 1 1 143
2 M8 1 1 1 1 1 1 95
1 M6 1 1 1 1 1 94
 1023 1020 963 816 747 695 446 303 Decim
al

𝒌
Group Efficacy =
𝒌+𝐞𝟎+𝐯

𝑘 = Number of operations in diagonal box

𝑒0 = No of exceptions

𝑣 = Number of voids

33
Group Efficacy =
33+22+7

Group Efficacy = 53.22 %

Observations and Improvements

Our current ROC method groups machines and parts with 52 % efficiency, leaving significant room

for improvement. This inefficiency stems from two issues:

1. Scattered Operations: Over half (48%) of operations aren't optimally grouped, indicating

poorly allocated parts to machines.

2. Limited Consideration: ROC only considers presence/absence of operations, potentially

missing crucial details about processing complexity.

• Refine Grouping: We can rearrange groups with similar processing patterns (both

machines and parts) to concentrate operations on the diagonal, leading to a more efficient

grouping.

• Explore New Algorithms: We can test the Column Interchange (CI) method and compare

its effectiveness to ROC. Additionally, exploring algorithms like K-means or hierarchical

clustering might reveal better grouping structures based on processing relationships.

• Incorporate Additional Data (if available): If we have data on processing times and setup

costs, we can leverage this information. Processing times can help prioritize efficient

groupings based on actual needs, while setup costs can prioritize minimizing machine

downtime through optimized part allocation.

By implementing these strategies, we can compare the effectiveness (Grouping Efficacy or GE)

of each approach. The method with the highest GE will likely lead to the most efficient machine-

part allocation!

Question no 02

Generate a 8×7 (8 machines and 7 parts) machine-part incidence matrix having at least 35

operations and then by applying any approach (Not the ones that we studied in class i.e. CIA and

ROC) block-diagonalize this matrix and calculate its Grouping Efficacy (GE). Can the outcome of

your selected approach beat the outcomes of ROC and CIA? Present a reasoning-based analysis

both in case of “Yes” and “No”.

Direct Cluster Algorithm Approach

In this method, Sum all 1’s of columns and rows and arrange in descending order.

Manufactures/Parts P1 P2 P3 P4 P5 P6 P7

M1 4
1 1 1 1 0 0 0
M2 4
1 1 1 0 1 0 0
M3 5
1 1 1 0 1 0 1
M4 2
1 0 0 1 0 0 0
M5 5
0 1 1 0 1 1 1
M6 5
0 0 1 1 1 1 1
M7 5
1 1 0 0 1 1 1
M8 5
0 1 0 1 1 1 1

5 6 5 4 6 4 6

Manufactures/Parts P6 P4 P7 P3 P1 P5 P2

M8 5
1 1 1 0 0 1 1
M7 5
1 0 1 0 1 1 1
M6 5
1 1 1 1 0 1 0
M5 5
1 0 1 1 0 1 1
M3 5
0 0 1 1 1 1 1
M2 4
0 0 0 1 1 1 1
M1 4
0 1 0 1 1 0 1
M4 2
0 1 0 0 1 0 0

4 4 5 5 5 6 6
Number of operations = 36
𝒌
Group Efficacy =
𝒌+𝐞𝟎+𝐯

23
Group Efficacy = = 56.09 %
23+12+6

Using Rank Order Clustering Algorithm

Manufactures/Parts P1 P2 P3 P4 P5 P6 P7

M1 1 1 1 1 0 0 0
M2 1 1 1 0 1 0 0
M3 1 1 1 0 1 0 1
M4 1 0 0 1 0 0 0
M5 0 1 1 0 1 1 1
M6 0 0 1 1 1 1 1
M7 1 1 0 0 1 1 1
M8 0 1 0 1 1 1 1

64 32 16 8 4 2 1

Manufactures/Parts P1 P2 P3 P4 P5 P6 P7

M1 1 1 1 1 0 0 0 120

M2 1 1 1 0 1 0 0 116

M3 1 1 1 0 1 0 1 117

M4 1 0 0 1 0 0 0 72

M5 0 1 1 0 1 1 1 55

M6 0 0 1 1 1 1 1 31

M7 1 1 0 0 1 1 1 103

M8 0 1 0 1 1 1 1 47
 64 32 16 8 4 2 1

Manufactures/Parts P1 P2 P3 P4 P5 P6 P7

128 M1 1 1 1 1 0 0 0 120

64 M3 1 1 1 0 1 0 1 117

32 M2 1 1 1 0 1 0 0 116

16 M7 1 1 0 0 1 1 1 103

8 M4 1 0 0 1 0 0 0 72

4 M5 0 1 1 0 1 1 1 55

2 M8 0 1 0 1 1 1 1 47

1 M6 0 0 1 1 1 1 1 31

248 246 229 139 119 23 87

64 32 16 8 4 2 1

Manufactures/Parts P1 P2 P3 P4 P5 P7 P6

128 M1 1 1 1 1 0 0 0 120

64 M3 1 1 1 0 1 1 0 117

32 M2 1 1 1 0 1 0 0 116

16 M7 1 1 0 0 1 1 1 103

8 M4 1 0 0 1 0 0 0 72

4 M5 0 1 1 0 1 1 1 55

2 M8 0 1 0 1 1 1 1 47

1 M6 0 0 1 1 1 1 1 31

248 246 229 139 119 87 23

𝒌
Group Efficacy =
𝒌+𝐞𝟎+𝐯
 𝟐𝟑 = 57.5 %
Group Efficacy = 𝟐𝟑+𝟏𝟐+𝟎𝟓

This section examines two new methods for organizing machines and parts within a machine-part

matrix. These methods, Direct Clustering and Rank Ordering, aim to achieve a block diagonal

structure, where processing relationships are clearly defined.

Can these new methods rival existing approaches?

It's unclear but likely no. The new methods are simpler, but established techniques like ROC (Rank

Order Clustering) and CIA (Column Interchangeability Algorithm) are more complex and often

achieve better results. These existing algorithms use iterative approaches to explore various

grouping options, potentially leading to higher "Grouping Efficiency" (GE).

Here's a breakdown of the new methods:

• Direct Clustering: They often provide more accurate results, especially when the
underlying structure of the data is well-defined.

• Rank Ordering: It can be useful when the exact values of data points are less important

than their relative positions.

The chosen example for evaluating these new methods might not be universally applicable. To get

a more conclusive answer, applying ROC and CIA to the same matrices used for the new methods

would be ideal.